331 SYSTEMIC FUNCTIONAL GRAMMAR AND CONSTRUCTION GRAMMAR Francis Y. LIN Alex X. PENG (School of Foreign Languages and Literatures / Beijing Normal University) ABSTRACT: Construction Grammar (CG) as developed by Fillmore, Goldberg and others is a recent development in syntactic theory, which has become more and more influential. Its central claim is that in a language there are a large number of grammatical units, called constructions, which are the basic forms for the speakers to express their meanings. Systemic Functional Grammar (SFG), put forward by Halliday, also pays great attention to how the speakers generate utterances and texts to convey their intended meanings. This paper explores the relationship between CG and SFG. It argues that the concept of constructions should be introduced in SFG and reflected in the transitivity network. It also suggests that main ideas from SFG be used in CG to describe language more adequately. The objective is to make SFG and CG better theories of language, by combining their strengths. KEY-WORDS: Construction Grammar, Systemic Functional Grammar, Way construction, Resultative construction 1. Introduction Construction Grammar (CG) as developed by Fillmore, Goldberg and others (Kay & Fillmore 1999; Goldberg 1995, 2006; Fillmore et al. forthcoming) is a recent development in syntactic theory, which has become more and more influential. Its central claim is that in a language there are a large number of grammatical units, called constructions, which are the basic forms for the speakers to express their meanings. Constructions are language-specific, as different languages have different means of expressing the same (or similar) meanings. Children on their way of mastering a full language acquire an increasing number of constructions, and become more skillful in using language. Proceedings 33rd International Systemic Functional Congress 2006 332 Systemic Functional Grammar (SFG), put forward by Halliday (Halliday 1994; Halliday & Matthiessen 2004), also pays great attention to how the speakers generate utterances and texts to convey their intended meanings. According to SFG, language has three metafunctions, namely, the ideational, interpersonal, and textual metafunctions. All the three metafunctions are reflected in a huge system network, which specifies all the meaning potentials. The system network consists of several subnetworks, such as the Transitivity network, the Thing network, and the Quality network. A network is made up by a number of the so-called ‘systems’, each consisting of a set of semantic features. To generate an utterance, the system network is traversed, certain semantic features are selected, and the relevant realization rules are fired (for details of sentence generation in SFG, see Fawcett, Tucker & Lin 1993; see also Mann & Matthiessen 1985). From the perspective of SFG, children gradually acquire a full system network, and use it to produce a large number of sentences. The similarity between CG and SFG is thus clear. Both model a speaker’s grammatical knowledge. In CG, grammatical knowledge is knowledge of a large number of constructions, which form a structured inventory of speakers’ knowledge of the conventions of their language (Langacker 1987: 63-66; Goldberg 2006: 18); and in SFG it is knowledge of a huge system network. But there are also many differences between the two theories. A major difference is in the process of sentence generation. According to CG, a speaker has a list of constructions at his disposal and he just selects one of them as the blueprint for making his utterance. For example, a speaker may need to select the so called Way construction, when needing to utter John whistled his way home or He belched his way out of the restaurant. But according to SFG, a speaker must traverse the system network, making various types of choices. In SFG, there is no explicit notion of constructions. And there is little research on how such sentences can be generated in the system network. It is important to incorporate the idea of sentence constructions into SFG. There are two major reasons for doing so. One is this. It is a fact that there are various constructions in a language. For SFG to describe languages faithfully and adequately, it must not ignore this fact and must somehow account for it. The second reason may have to do with technicality. The Proceedings 33rd International Systemic Functional Congress 2006 333 constructions in a language are large in number. For example, in English, apart from the Way construction discussed above, there are other constructions such as the Resultative construction (e.g. Mary wiped the table clean), the the X-er, the Y-er construction (e.g. The more you read it, the better you will understand it), to name a few. It would be very difficult to incorporate all such constructions into the existing system network. The organization of the system network needs to be carefully rethought in order to account for such sentence constructions, which are facts of language. This paper aims to combine the strengths of SFG and CG, especially by incorporating the idea of constructions from CG into SFG. Section 2 explicates the idea of constructions. Section 3 compares SFG and CG, pointing out their similarities in treating simple sentence constructions. Section 4 compares how SFG and CG deal with complex sentence constructions. Section 5 suggests a way of incorporating the idea of constructions into SFG. Section 6 concludes this paper and discusses some related issues. 2. Constructions According to CG, constructions are “conventionalized parings of form and function” (Goldberg 2006: 3). Constructions vary in size and complexity, ranging from morphemes or words, through idioms, phrases, to sentences (Goldberg 2006: 5). In this paper we only concentrate on constructions at the sentence level. We distinguish between two types of sentence constructions: ‘simple constructions’ and ‘complex constructions’. A simple sentence construction consists of at least a main verb V. It often also has a subject; it may also have an object, which may be a thing, a location, an attribute etc. So, typical simple sentence constructions are of the form: S V, and S V O. And typical simple sentences are John smiles, Peter kicked a ball, His house is in London, She is very pretty, etc. Simple constructions are closely related to the valency structure of the verbs. But there are also constructions which are not determined by the valency structures. For example, the verb wipe has the valency structure Proceedings 33rd International Systemic Functional Congress 2006 334 X wipe Y. But we can say She wiped the table clean, which is of the construction X wipe Y ADJ. Similarly, there are many other such complex constructions, such as the Way construction, the the X-er, the Y-er construction, and so on. We will first compare the analyses of simple constructions in SFG and in CG, and then the treatments of simple constructions in the two theories. 3. Simple constructions: SFG and CG contrasted 3.1. The SFG treatment In generating a simple sentence (e.g. Peter kicked a ball), the transitivity network is traversed first. The result is a skeleton sentence, e.g.: Clause S/Actor V O/Goal kicked Figure 1: A sample skeleton sentence The subject S (e.g. Peter), which is the Actor of the kicking process, will be generated by traversing the thing network. So will be the object (e.g. a ball), which is the Goal of the kicking the process. (See Fawcett, Tucker & Lin 1993; Fawcett 2000) This skeleton sentence is a mixture of semantic structure and syntactic structure, which can be depicted in Table 1 below (cf.: Halliday 1994; Halliday & Matthiessen 2004): Proceedings 33rd International Systemic Functional Congress 2006 335 Sentence Peter kicked a ball Semantic Structure Actor Process Goal Syntactic Structure Subject Main Verb Object Table 1: SFG analysis of Peter kicked a ball 3.2. The CG treatment CG claims that the basic units of language are constructions. In order to generate a simple sentence, a simple construction is first selected, which has a corresponding semantics. For example, to generate Peter kicked a ball, the construction S V O is selected, which has the semantics <Agent Predicate Goal>. The relationship between the syntax and semantics of the sentence can be depicted as below (see Goldberg 1995: 117): Semantics KICKING < agent patient> Syntax V SUBJ OBJ Figure 2: CG analysis of Peter kicked a ball The construction S V O is not yet a concrete sentence. In order to produce Peter kicked a ball, S will need to be filled by Peter, V by kicked, and O by a ball. The production of Peter, kicked, and a ball will involve the selection of certain semantic features, such as [name_for_man], [hitting_with_foot], [singular, round_object], respectively. See Fillmore et al. (forthcoming). 3.3. Comparison between CG and SFG The above two subsections explicated the treatment of simple sentences in SFG and that in CG. We are now in a position to compare them. Take the sentence Peter kicked a ball for illustration. Proceedings 33rd International Systemic Functional Congress 2006 336 Firstly, SFG and CG give the same semantic analysis, except with some terminological differences. SFG and CG agree that the sentence Peter kicked a ball describes a process (or a predicate), which involves two participant roles: an Actor (or Agent), and a Goal (Patient). Secondly, SFG and CG provide the same syntactic analysis, except with some terminological differences. Both regard the sentence as consisting of a subject and an object. Thirdly, SFG and CG provide the same linking relationship between the semantic structure and the syntactic structure. Both think that the subject plays the role Actor (or Agent), and object the role Goal (Patient). Fourthly, both SFG and CG produce an incomplete sentence first, and then fills the empty slots with concrete words or phrases.